Abstract
This study presents an application of the Particle Swarm Optimization (PSO) algorithm to the parameter optimization of rainfall-runoff models. Six global optimization algorithms, the shuffled complex evolution method (SCE-UA), modified SCE-UA, modified SCE-UA with initial value, PSO, modified PSO and modified PSO with initial value, were applied to parameter optimization on four kinds of series tank models. Performance comparison of there algorithms was evaluated and it can be concluded that SCE-UA and PSO show comparable performance in most cases. In addition, PSO is more effective than SCE-UA under the following conditions. 1) the model has large number of parameters, 2) the model has wide range of parameters, 3) calibration period is too short, 4) observation data contains large uncertainty. The modified PSO with initial value shows the most effective and stable performance.
